A novel approach to robust motion control of electrical drives with model order uncertainty

摘要

A novel approach to the control of plants with model order uncertainty as well as parametric errors and externaldisturbances is presented, which yields a specified settling time of the step response with zero overshoot. The method is applied to amotion control system employing a permanent magnet synchronous motor. A single controller is designed to cater for mechanicalloads that may exhibit significant vibration modes. The order of the complete controlled system (i.e., the plant) will therefore dependon the number of significant vibration modes. The controller is of the cascade structure, comprising an inner drive speed control loopand an outer position control loop. The main contribution of the paper is a completely new robust control strategy for plants withmodel order uncertainty, which is used in the outer position control loop. Its foundations lie in sliding mode control, but the set ofoutput derivatives fed back extend to a maximum order depending on the maximum likely rank of the plant, rather than its knownrank. In cases where the maximum order of output derivative exceeds the plant rank, in theory, virtual states are created that raise theorder of the closed-loop system while retaining the extreme robustness properties of sliding mode control. Algebraic loops (causedby zero or negative rank of the open-loop system) are avoided by embodying filtering with a relatively short time constant in theoutput derivative approximations. The speed control loop is also new. Although it is based on the forced dynamic vector controlprinciple, already developed by the author and co-researchers for drives with current fed inverters, for the first time, a version forvoltage fed inverters is presented with a view to future implementation of space vector modulation to improve the smoothness of thestator current waveforms. The new forced dynamic control law requires an estimate of the load torque and its first derivative and aspecial observer is presented for this purpose. An initial evaluation of the method is made by considering three plants with differentorders and ranks, the first being the unloaded drive, the second being the drive controlling the motor rotor angle with a mass-springload attached and the third being the drive controlling the load mass angle of the same attached mass-spring load. The simulationsindicate that the control system does indeed yield robustness including plant order uncertainty and further investigations, boththeoretical and experimental, are recommended.